Antithrombotic outcome trials in acute coronary syndromes: seeking the optimal balance between safety and efficacy†

An intricate balance exists between the safety and efficacy of antithrombotic therapies in acute coronary syndrome (ACS). Safety is largely defined by bleeding risks. Many bleeding definitions are used in clinical trials, and conclusions may differ depending on the definition applied. Bleeding risks are typically dose related, making selection of the optimal dose (or preferably, doses) to advance from phase II to III testing a critical step in drug development.1 Efficacy is defined in terms of thrombotic event reduction and is commonly assessed using composite endpoints. However, composite endpoints create interpretive challenges within and across trials, particularly when endpoint definitions vary substantially. The trade-off between benefits (reduction in thrombotic events) and risks (increase in bleeding) is important to patients and clinicians; thus, it is imperative that data are presented clearly to facilitate informed decision making. Antithrombotic drug development for ACS is a challenging area of cardiovascular medicine, a topic discussed during the 7th and 8th Global Cardiovascular Clinical Trialists (CVCT) forums in December 2010 and 2011. This manuscript highlights key points from those discussions. Data from phase II trials are used to determine dose selection for phase III. However, it is not uncommon for phase II and III results to differ substantially (Tables 1–3). In the phase II studies of vorapaxar, prasugrel, and ticagrelor, there was no evidence of a statistically significant increase in bleeding when compared with standard therapy or placebo. However, in the larger phase III studies, an increased risk of bleeding not associated with coronary artery bypass graft (CABG) surgery was observed (Tables 2 and 3). In the phase II studies of apixaban and rivaroxaban, dose-related increases in bleeding were observed leading to the selection of lower doses for phase III. Despite the dose reduction, bleeding was still higher with these drugs when compared with placebo in the pivotal trials (Tables 2 and 3).2,3 Overview of recent acute coronary syndrome phase II and phase III trials Overview of recent acute coronary syndrome phase II and phase III trials Clinical and Bleeding outcomes in recent trials of antithrombotic therapy NR, not reported; CTVT, clinical target vessel thrombosis. aSee Table 1 for definition of primary efficacy and bleeding endpoints. bOnly b.i.d. dosing arms shown due to space limitations. cAll P-values are for the combined rivaroxaban dosing groups vs. placebo. Clinical and Bleeding outcomes in recent trials of antithrombotic therapy NR, not reported; CTVT, clinical target vessel thrombosis. aSee Table 1 for definition of primary efficacy and bleeding endpoints. bOnly b.i.d. dosing arms shown due to space limitations. cAll P-values are for the combined rivaroxaban dosing groups vs. placebo. Clinical and bleeding outcomes in recent trials of antithrombotic therapy NR, not reported. aSee Table 1 for definition of primary efficacy and bleeding endpoints. bOnly primary PCI cohort data shown. Clinical and bleeding outcomes in recent trials of antithrombotic therapy NR, not reported. aSee Table 1 for definition of primary efficacy and bleeding endpoints. bOnly primary PCI cohort data shown. These experiences illustrate the difficulties with choosing a dose for phase III based on phase II data. These trials were large relative to many phase II trials in other diseases, but follow-up was short and event rates low, thereby decreasing their power to be informative regarding either safety or efficacy outcomes. Small numbers (and limited power) are misleading, and conclusions drawn under these conditions are exposed to error. Most of the phase II studies showed relative increases in bleeding with the novel antithrombotic agents, and these increases became significant in the larger phase III studies that had greater power to detect such differences at typical levels of statistical significance. The phase II and phase III hazard ratios and 95% confidence intervals become more precise in the phase III trials, with more patients and longer follow-up (Figures 1 and 2). Comparison of hazard ratios and 95% CI for primary efficacy outcomes in phase II and III antithrombotic acute coronary syndrome trials. Comparison of hazard ratios and 95% CI for bleeding outcomes in phase II and III antithrombotic acute coronary syndrome trials Supplementary material online, Figures S1 and S2. Endpoint selection in phase II and III may also contribute to the different observations within these trials. In both the apixaban and rivaroxaban phase II studies, the efficacy endpoint was a combination of cardiovascular death, myocardial infarction (MI), recurrent ischaemia and ischaemic stroke, while in phase III, the endpoint was cardiovascular death, MI and ischaemic stroke (Table 1). The inclusion of more (usually ‘softer’) components may be misleading if the study drug has a larger effect on the lesser important components (e.g. recurrent ischaemia) than on the components measured in phase III (e.g. cardiovascular death). It may also be misleading if the softer components add ‘noise’, i.e. additional events that are unaffected by the study agent (e.g. in TRACER,4 no significant difference was noted in the primary endpoint of cardiovascular death, MI, stroke, recurrent ischaemia with hospitalization or urgent coronary revascularization, but a statistically significant difference was observed in the composite of cardiovascular death, MI or stroke).4 Finally, phase II design is crucial if it is used to plan phase III. The ATLAS ACS TIMI 46 trial was large and completed as planned.5 Two doses were carried forward from phase II to the phase III ATLAS ACS 2-TIMI 51.2 In comparison, the phase II APPRAISE trial was smaller, and enrolment in the higher doses was stopped because of excess bleeding.6 The single dosing regimen chosen for APPRAISE-2 was not one of the doses studied in the phase II trial.3 Adaptive designs to explore safety (bleeding) have been proposed to overcome the limitations of phase III dose selection.7 Using an adaptive approach, seamless phase II/III studies are designed with multiple doses, and the number of dosing groups adaptively decreases during the study according to a pre-specified analysis plan at one or more unblinded interim analyses. Doses with an unfavourable risk/benefit ratio are dropped. The sample size of individual dosing groups or the overall study may be adjusted using the accruing efficacy and safety data.8,9 Another potentially useful adaptive design is sample size re-estimation based on unblinded interim results, whereby if the interim treatment difference is promising (but not as large as that originally powered for) the trial size is increased accordingly.10 Appropriate methods to control type I error for safety and efficacy, blinding of the steering committee, processes for communication between the steering and data safety monitoring committees and implementation of adaptive changes are challenges associated with these designs that remain unresolved.11 Adaptive designs will only be useful if they are accepted by the regulatory agencies. Both European and US regulators have addressed this topic, but they have not announced a final position.8,9 These designs hold promise for increasing the efficiency of clinical trials, but more research and experience are needed before their widespread adoption.7–9 Major bleeding is an immediate safety outcome and an independent predictor of subsequent mortality among ACS patients, although whether this relationship is causal or simply identifies higher-risk patients is unknown.12,13 The impact of a bleeding event on mortality persists during the long-term follow-up, in contrast to the influence of a recurrent MI on mortality, which is mainly short term.13,14 Bleeding events, even those classified as minor, may lead to discontinuation of antithrombotic therapy, which increases the risk of recurrent ischaemic events.15 The use of transfusions also carries long-term risks.16 Importantly, bleeding events prolong hospital stays and increase costs. However, there is a marked variability in bleeding definitions used across trials (Supplementary material online, Appendix), which severely limits data interpretation. The components of bleeding endpoints are not equal from the patient perspective (e.g. an intracranial bleed is more severe than bleeding related to instrumentation) or from the standpoint of effectiveness research. Uniform and comprehensive reporting of bleeding events and associated outcomes facilitates event-level data comparisons between agents and across trials. Relevant reporting parameters include, but are not limited to, bleed location, amount of blood loss, need for interventions and resulting disability. In an effort to achieve consistency, the Standardized Data Collection for Cardiovascular Trials Initiative has reviewed existing bleeding definitions and posted recommendations for public comment, but the final definitions have not been announced.17 The Bleeding Academic Research Consortium (BARC) has also proposed a 5-point scoring system for bleeding (Supplementary material online, Appendix).18 Each of these definitions aims to quantify bleeding severity, but they use different criteria. Thus, various definitions applied to the same data will result in calculation of different bleeding rates. It is apparent that although efforts are being made towards adoption of consistent definitions, many options to define bleeding will be available for use in future clinical trials. Therefore, regardless of the final definitions, study results should be presented such that bleeding outcomes can be interpreted and compared across trials to achieve a comprehensive understanding of bleeding risk. The optimal antithrombotic regimen should achieve the greatest reduction in thrombotic events at the lowest bleeding risk. This balance is often difficult to achieve because efficacy and safety are related to the same pharmacodynamic effect. It is further complicated by the limited ability to compare data across studies. These concepts are illustrated using several relevant clinical trials. In the Trial to Assess Improvement in Therapeutic Outcomes by Optimizing Platelet Inhibition with Prasugrel-Thrombolysis in Myocardial Infarction (TRITON-TIMI 38) trial, TIMI major bleeding not related to CABG was increased with prasugrel when compared with clopidogrel (Supplementary material online, Figure S1a, Tables 2 and 3).19 Further analysis revealed a statistically significant increased risk of life-threatening bleeding with prasugrel that was driven by fatal and spontaneous non-fatal events.19 In the Study of Platelet Inhibition and Patient Outcomes (PLATO) trial, there was no difference in major bleeding between groups. However, this included and was dominated by bleeding related to CABG. When the bleeding criteria used in the TRITON-TIMI 38 study were applied to PLATO (i.e. excluding CABG-related bleeding), statistically significant increases in bleeding and fatal intracranial bleeding were noted with ticagrelor (Supplementary material online, Figure S1a). In both trials, death from vascular causes, MI or stroke was lower in patients randomized to prasugrel or ticagrelor when compared with clopidogrel (Supplementary material online, Figure S1b, Tables 2 and 3).19,20 Myocardial infarction (both spontaneous and procedure related)21 was the main contributor to the primary endpoint in both studies (Tables 2 and 3). In ATLAS ACS 2-TIMI 51, rivaroxaban was associated with absolute increases in TIMI major non-CABG bleeding of 1.2% (2.5 mg) and 1.8% (5 mg) compared with placebo, and a 0.2% (2.5 mg b.i.d.) and 0.5% (5 mg b.i.d.) absolute increase in intracranial bleeding. Absolute reductions of 1.8% in the primary composite endpoint (cardiovascular death, MI, or stroke), 0.8% in all-cause mortality and 1.1% in MI were also observed2 (Tables 2 and 3). The exact balance between risks and benefits is uncertain due to incomplete follow-up in some patients, which caused the FDA to withhold the approval of rivaroxaban for an ACS indication pending further data. Although definite conclusions cannot be drawn from post hoc analyses of individual components of bleeding or thrombotic outcomes that occur with a low frequency, the importance of evaluating the totality of evidence becomes clear. Clinically meaningful increases or decreases in specific thrombotic or bleeding events may be masked by an overall neutral composite result (e.g. overall bleeding composite is neutral but life-threatening or intracranial bleeding is increased). Similarly, a composite may be driven by a factor with a lower degree of clinical importance (e.g. small non-fatal MI), without effect on more important events, particularly mortality. For this reason, net clinical benefit composite endpoints are problematic, since each outcome is considered clinically equivalent, while this assumption is obviously incorrect. Similar to other types of composite endpoints, net clinical benefit composites can hide components that go in divergent directions. As with all composites, the components should be reported individually. New methodologies are being developed to deal with these limitations of composite endpoints. One major limitation is that time-to-first event analyses do not consider subsequent events. Less clinically important events often occur first and count towards the endpoint (e.g. non-fatal MI), whereas subsequent, often more severe events, are disregarded. The win ratio approach to analysing composite endpoints gives greater priority to the more clinically important components, such as mortality.22 With this method, the most important component of the composite is determined. For example, in a composite of cardiovascular death, MI or stroke, cardiovascular death would be the most important component, followed by stroke, and then MI. Patients in the active treatment and control (or standard) treatment are matched using a risk score or risk stratification, and compared to determine which patient experienced a cardiovascular death first. If neither experienced a death, then they would be compared to determine who experienced a stroke first, and so on. The ‘win ratio’ is the number of matched pairs where patients on standard therapy had worse outcomes. An alternative unmatched ‘win ratio’ approach compares every active with every control patient using the same principles; this may be preferred if the basis of matching cannot be clearly defined in advance. Although more research is needed to validate this approach and to determine its acceptance by regulatory bodies, it is encouraging that more robust and insightful methods of analysing composite endpoints are on the horizon. New therapies for ACS must be tested against evidence-based therapy as recommended in international guidelines (aspirin and a P2Y12 inhibitor, either ticagrelor, prasugrel, or clopidogrel depending on the clinical situation).23,24 The additional treatment effects are likely to be small, while the additive bleeding risks may be substantial. Very large sample sizes will be required to detect treatment effects, and the net clinical benefit may be small if bleeding risks are substantially increased. The polypharmacy of double and triple drug regimens carries important implications from the standpoints of adherence, cost, side effects, and the balance of risks and benefits. Consideration should be given to trials that evaluate novel single drugs or combinations of drugs (e.g. can newer drugs replace aspirin or P2Y12 inhibitors) rather than triple or quadruple therapy. A direct comparison of prasugrel and ticagrelor would be of interest to assess whether the observed differences in the TRITON and PLATO trials are real and reproducible. Research that aims to quantify the contribution of individual drugs would be clinically meaningful. For example, to what extent does an individual drug reduce thrombotic events and increase bleeding risk? Such research might lead to modification of current guidelines and a shift in clinical equipoise with regard to what constitutes ‘standard therapy’ for future trials. However, careful attention needs to be placed on the ethics of such designs to ensure that the best standard of care is not compromised. Some of the points raised in this review may appear semantic, but they become clinically relevant when one applies the results of trials to clinical practice. What level of bleeding risk will patients and physicians be willing to accept to achieve which level and type of thrombotic event reduction, particularly when non-fatal MI is driven by biomarker positivity? Is thrombotic event reduction alone (reduction of MI and stent thrombosis) in the of benefit on What of effect on cardiovascular or all-cause mortality should be to bleeding These also patients with drugs associated with a significant bleeding risk for clinical benefits. These remain but are critical concepts that The development of antithrombotic agents for the treatment of ACS and other cardiovascular is a and Optimizing the of ischaemic events and bleeding risks are to patient outcomes. This target is often because efficacy and safety are in ACS have event rates such that of of patients are required to achieve power to detect small treatment This the use of composite efficacy and safety endpoints, so that numbers of events can in a and physicians are then with the challenges in composite effects that may differ among the individual components of the results that are driven by a important or softer in efficacy and safety endpoint and safety outcomes. The proposed multiple for future research to some of these challenges (Table additional analyses of existing and an among and such as in are needed to these challenges and to and therapies to the of ACS for future research reporting criteria for bleeding bleeding on bleeding location, of and outcome If the of such then the should be available as an Supplementary material online, in the clinical trial or of composite endpoints should be reported MI type and The win ratio or other novel methodologies should be in completed trial to experience with the and to determine whether it will be an for standard analyses. Adaptive designs should be considered for phase II trials to more robust data to phase III. should on with regard to methodologies for data analysis and with type I error. the optimal of risk vs. benefit with regard to antithrombotic therapy, more research is needed to determine the between different levels of bleeding and subsequent clinical outcomes and the between thrombotic events and subsequent clinical outcomes. research efforts should to quantify the relative contribution of individual antithrombotic therapies to ischaemic event reduction and bleeding in an effort to what should ‘standard of for future trials. Research evaluating patient reported outcomes should on the level of risk vs. benefit from the as as the impact of triple drug therapy on the of reporting criteria for bleeding bleeding on bleeding location, of and outcome If the of such then the should be available as an Supplementary material online, in the clinical trial or of composite endpoints should be reported MI type and The win ratio or other novel methodologies should be in completed trial to experience with the and to determine whether it will be an for standard analyses. Adaptive designs should be considered for phase II trials to more robust data to phase III. should on with regard to methodologies for data analysis and with type I error. the optimal of risk vs. benefit with regard to antithrombotic therapy, more research is needed to determine the between different levels of bleeding and subsequent clinical outcomes and the between thrombotic events and subsequent clinical outcomes. research efforts should to quantify the relative contribution of individual antithrombotic therapies to ischaemic event reduction and bleeding in an effort to what should ‘standard of for future trials. Research evaluating patient reported outcomes should on the level of risk vs. benefit from the as as the impact of triple drug therapy on the of for future research reporting criteria for bleeding bleeding on bleeding location, of and outcome If the of such then the should be available as an Supplementary material online, in the clinical trial or of composite endpoints should be reported MI type and The win ratio or other novel methodologies should be in completed trial to experience with the and to determine whether it will be an for standard analyses. Adaptive designs should be considered for phase II trials to more robust data to phase III. should on with regard to methodologies for data analysis and with type I error. the optimal of risk vs. benefit with regard to antithrombotic therapy, more research is needed to determine the between different levels of bleeding and subsequent clinical outcomes and the between thrombotic events and subsequent clinical outcomes. research efforts should to quantify the relative contribution of individual antithrombotic therapies to ischaemic event reduction and bleeding in an effort to what should ‘standard of for future trials. Research evaluating patient reported outcomes should on the level of risk vs. benefit from the as as the impact of triple drug therapy on the of reporting criteria for bleeding bleeding on bleeding location, of and outcome If the of such then the should be available as an Supplementary material online, in the clinical trial or of composite endpoints should be reported MI type and The win ratio or other novel methodologies should be in completed trial to experience with the and to determine whether it will be an for standard analyses. Adaptive designs should be considered for phase II trials to more robust data to phase III. should on with regard to methodologies for data analysis and with type I error. the optimal of risk vs. benefit with regard to antithrombotic therapy, more research is needed to determine the between different levels of bleeding and subsequent clinical outcomes and the between thrombotic events and subsequent clinical outcomes. research efforts should to quantify the relative contribution of individual antithrombotic therapies to ischaemic event reduction and bleeding in an effort to what should ‘standard of for future trials. Research evaluating patient reported outcomes should on the level of risk vs. benefit from the as as the impact of triple drug therapy on the of Supplementary material is available at European This was from during the 7th and 8th Global Cardiovascular Clinical Trialists (CVCT) in in December 2010 and 2011. was by the Clinical and of and by an from a in had no in or approval of the manuscript for of and research from and from The and Research or research and from The and of for and research from The and for for for a clinical trial of from Data Safety for APPRAISE studies. development on and event on trial on development on trial of of of The of to 2010 and and related to of this The were or the topic of this manuscript at the December 2010 Global Cardiovascular Clinical Trialists